OV - exec - Loop 94

	Loop Id: 94	Module: exec	Source: timestep.c:86-96	Coverage: 0.04%

0x40ec50 INC	%RAX

0x40ec53 ADD	$0x40,%ESI

0x40ec56 CMP	%RCX,%RAX

0x40ec59 JE	40ec1b

0x40ec5b MOV	(%RDX,%RAX,4),%R11D

0x40ec5f TEST	%R11D,%R11D

0x40ec62 JLE	40ec50

0x40ec64 MOV	%ESI,%R9D

0x40ec67 MOV	0x20(%RBX),%R8

0x40ec6b MOV	0x28(%RBX),%R10

0x40ec6f MOV	0x10(%R8),%R15

0x40ec73 MOV	0x18(%R8),%RDI

0x40ec77 MOV	0x20(%R8),%R8

0x40ec7b CMP	$0x1,%R11D

0x40ec7f JE	40ed4c

0x40ec85 LEA	(%R9,%R9,2),%R12

0x40ec89 LEA	0x28(,%R12,8),%R12

0x40ec91 MOV	%R11D,%R13D

0x40ec94 AND	$0x7ffffffe,%R13D

0x40ec9b NOPL	(%RAX,%RAX,1)

  (95) 0x40eca0 MOVSXD	(%R15,%R9,4),%R14

  (95) 0x40eca4 SAL	$0x4,%R14

  (95) 0x40eca8 VMOVSD	-0x28(%R8,%R12,1),%XMM1

  (95) 0x40ecaf VDIVSD	0x8(%R10,%R14,1),%XMM0,%XMM2

  (95) 0x40ecb6 VFMADD213SD	-0x28(%RDI,%R12,1),%XMM2,%XMM1

  (95) 0x40ecbd VMOVSD	%XMM1,-0x28(%RDI,%R12,1)

  (95) 0x40ecc4 VMOVSD	-0x20(%R8,%R12,1),%XMM1

  (95) 0x40eccb VFMADD213SD	-0x20(%RDI,%R12,1),%XMM2,%XMM1

  (95) 0x40ecd2 VMOVSD	%XMM1,-0x20(%RDI,%R12,1)

  (95) 0x40ecd9 VMOVSD	-0x18(%R8,%R12,1),%XMM1

  (95) 0x40ece0 VFMADD213SD	-0x18(%RDI,%R12,1),%XMM2,%XMM1

  (95) 0x40ece7 VMOVSD	%XMM1,-0x18(%RDI,%R12,1)

  (95) 0x40ecee MOVSXD	0x4(%R15,%R9,4),%R14

  (95) 0x40ecf3 SAL	$0x4,%R14

  (95) 0x40ecf7 VMOVSD	-0x10(%R8,%R12,1),%XMM1

  (95) 0x40ecfe VDIVSD	0x8(%R10,%R14,1),%XMM0,%XMM2

  (95) 0x40ed05 VFMADD213SD	-0x10(%RDI,%R12,1),%XMM2,%XMM1

  (95) 0x40ed0c VMOVSD	%XMM1,-0x10(%RDI,%R12,1)

  (95) 0x40ed13 VMOVSD	-0x8(%R8,%R12,1),%XMM1

  (95) 0x40ed1a VFMADD213SD	-0x8(%RDI,%R12,1),%XMM2,%XMM1

  (95) 0x40ed21 VMOVSD	%XMM1,-0x8(%RDI,%R12,1)

  (95) 0x40ed28 VMOVSD	(%R8,%R12,1),%XMM1

  (95) 0x40ed2e VFMADD213SD	(%RDI,%R12,1),%XMM2,%XMM1

  (95) 0x40ed34 VMOVSD	%XMM1,(%RDI,%R12,1)

  (95) 0x40ed3a ADD	$0x2,%R9

  (95) 0x40ed3e ADD	$0x30,%R12

  (95) 0x40ed42 ADD	$-0x2,%R13D

  (95) 0x40ed46 JNE	40eca0

0x40ed4c TEST	$0x1,%R11B

0x40ed50 JE	40ec50

0x40ed56 MOVSXD	(%R15,%R9,4),%R11

0x40ed5a SAL	$0x4,%R11

0x40ed5e SAL	$0x3,%R9

0x40ed62 LEA	(%R9,%R9,2),%R9

0x40ed66 VMOVSD	(%R8,%R9,1),%XMM1

0x40ed6c VDIVSD	0x8(%R10,%R11,1),%XMM0,%XMM2

0x40ed73 VFMADD213SD	(%RDI,%R9,1),%XMM2,%XMM1

0x40ed79 VMOVSD	%XMM1,(%RDI,%R9,1)

0x40ed7f VMOVSD	0x8(%R8,%R9,1),%XMM1

0x40ed86 VFMADD213SD	0x8(%RDI,%R9,1),%XMM2,%XMM1

0x40ed8d VMOVSD	%XMM1,0x8(%RDI,%R9,1)

0x40ed94 VMOVSD	0x10(%R8,%R9,1),%XMM1

0x40ed9b VFMADD213SD	0x10(%RDI,%R9,1),%XMM2,%XMM1

0x40eda2 VMOVSD	%XMM1,0x10(%RDI,%R9,1)

0x40eda9 JMP	40ec50

/home/eoseret/qaas_runs_GNR/173-814-1858/intel/CoMD/build/CoMD/CoMD/src-openmp/timestep.c: 86 - 96

--------------------------------------------------------------------------------

86:    for (int iBox=0; iBox<nBoxes; iBox++)

87:    {

88:       for (int iOff=MAXATOMS*iBox,ii=0; ii<s->boxes->nAtoms[iBox]; ii++,iOff++)

89:       {

90:          int iSpecies = s->atoms->iSpecies[iOff];

91:          real_t invMass = 1.0/s->species[iSpecies].mass;

92:          s->atoms->r[iOff][0] += dt*s->atoms->p[iOff][0]*invMass;

93:          s->atoms->r[iOff][1] += dt*s->atoms->p[iOff][1]*invMass;

94:          s->atoms->r[iOff][2] += dt*s->atoms->p[iOff][2]*invMass;

95:       }

96:    }

Coverage (%)	Name	Source Location	Module
►100.00+	__kmp_invoke_microtask		libiomp5.so
○	__kmp_invoke_task_func		libiomp5.so

Path /

Metric	Value
CQA speedup if no scalar integer	1.71
CQA speedup if FP arith vectorized	2.31
CQA speedup if fully vectorized	3.42
CQA speedup if no inter-iteration dependency	NA
CQA speedup if next bottleneck killed	1.46
Bottlenecks	micro-operation queue,
Function	advancePosition.extracted
Source	timestep.c:86-96
Source loop unroll info	NA
Source loop unroll confidence level	NA
Unroll/vectorization loop type	NA
Unroll factor	NA
CQA cycles	6.83
CQA cycles if no scalar integer	4.00
CQA cycles if FP arith vectorized	2.95
CQA cycles if fully vectorized	2.00
Front-end cycles	6.83
P0 cycles	3.50
P1 cycles	3.00
P2 cycles	4.67
P3 cycles	4.67
P4 cycles	1.50
P5 cycles	3.00
P6 cycles	3.50
P7 cycles	1.50
P8 cycles	1.50
P9 cycles	1.50
P10 cycles	3.00
P11 cycles	4.67
DIV/SQRT cycles	4.00
Inter-iter dependencies cycles	NA
FE+BE cycles (UFS)	NA
Stall cycles (UFS)	NA
Nb insns	37.00
Nb uops	37.00
Nb loads	14.00
Nb stores	3.00
Nb stack references	0.00
FLOP/cycle	1.02
Nb FLOP add-sub	0.00
Nb FLOP mul	0.00
Nb FLOP fma	3.00
Nb FLOP div	1.00
Nb FLOP rcp	0.00
Nb FLOP sqrt	0.00
Nb FLOP rsqrt	0.00
Bytes/cycle	18.73
Bytes prefetched	0.00
Bytes loaded	104.00
Bytes stored	24.00
Stride 0	NA
Stride 1	NA
Stride n	NA
Stride unknown	NA
Stride indirect	NA
Vectorization ratio all	0.00
Vectorization ratio load	0.00
Vectorization ratio store	0.00
Vectorization ratio mul	NA
Vectorization ratio add_sub	0.00
Vectorization ratio fma	0.00
Vectorization ratio div_sqrt	0.00
Vectorization ratio other	0.00
Vector-efficiency ratio all	10.55
Vector-efficiency ratio load	12.50
Vector-efficiency ratio store	12.50
Vector-efficiency ratio mul	NA
Vector-efficiency ratio add_sub	6.25
Vector-efficiency ratio fma	12.50
Vector-efficiency ratio div_sqrt	12.50
Vector-efficiency ratio other	7.50

Metric	Value
CQA speedup if no scalar integer	1.71
CQA speedup if FP arith vectorized	2.31
CQA speedup if fully vectorized	3.42
CQA speedup if no inter-iteration dependency	NA
CQA speedup if next bottleneck killed	1.46
Bottlenecks	micro-operation queue,
Function	advancePosition.extracted
Source	timestep.c:86-96
Source loop unroll info	NA
Source loop unroll confidence level	NA
Unroll/vectorization loop type	NA
Unroll factor	NA
CQA cycles	6.83
CQA cycles if no scalar integer	4.00
CQA cycles if FP arith vectorized	2.95
CQA cycles if fully vectorized	2.00
Front-end cycles	6.83
P0 cycles	3.50
P1 cycles	3.00
P2 cycles	4.67
P3 cycles	4.67
P4 cycles	1.50
P5 cycles	3.00
P6 cycles	3.50
P7 cycles	1.50
P8 cycles	1.50
P9 cycles	1.50
P10 cycles	3.00
P11 cycles	4.67
DIV/SQRT cycles	4.00
Inter-iter dependencies cycles	NA
FE+BE cycles (UFS)	NA
Stall cycles (UFS)	NA
Nb insns	37.00
Nb uops	37.00
Nb loads	14.00
Nb stores	3.00
Nb stack references	0.00
FLOP/cycle	1.02
Nb FLOP add-sub	0.00
Nb FLOP mul	0.00
Nb FLOP fma	3.00
Nb FLOP div	1.00
Nb FLOP rcp	0.00
Nb FLOP sqrt	0.00
Nb FLOP rsqrt	0.00
Bytes/cycle	18.73
Bytes prefetched	0.00
Bytes loaded	104.00
Bytes stored	24.00
Stride 0	NA
Stride 1	NA
Stride n	NA
Stride unknown	NA
Stride indirect	NA
Vectorization ratio all	0.00
Vectorization ratio load	0.00
Vectorization ratio store	0.00
Vectorization ratio mul	NA
Vectorization ratio add_sub	0.00
Vectorization ratio fma	0.00
Vectorization ratio div_sqrt	0.00
Vectorization ratio other	0.00
Vector-efficiency ratio all	10.55
Vector-efficiency ratio load	12.50
Vector-efficiency ratio store	12.50
Vector-efficiency ratio mul	NA
Vector-efficiency ratio add_sub	6.25
Vector-efficiency ratio fma	12.50
Vector-efficiency ratio div_sqrt	12.50
Vector-efficiency ratio other	7.50

Path /

Average path: Display a virtual path defined by average values of all real paths

Function	advancePosition.extracted
Source file and lines	timestep.c:86-96
Module	exec

The loop is defined in /home/eoseret/qaas_runs_GNR/173-814-1858/intel/CoMD/build/CoMD/CoMD/src-openmp/timestep.c:86-96.

Warnings:

Non-innermost loop: analyzing only self part (ignoring child loops).
Ignoring paths for analysis
Too many paths. Rerun with max-paths=5
RecMII not computed since number of paths is unknown or > max_paths
Streams not analyzed since number of paths is unknown or > max_paths

Try to simplify control and/or increase the maximum number of paths per function/loop through the 'max-paths-nb' option.

This loop has 5 execution paths.

The presence of multiple execution paths is typically the main/first bottleneck.
Try to simplify control inside loop: ideally, try to remove all conditional expressions, for example by (if applicable):

hoisting them (moving them outside the loop)
turning them into conditional moves, MIN or MAX

Ex: if (x<0) x=0 => x = (x<0 ? 0 : x) (or MAX(0,x) after defining the corresponding macro)

gain
potential
hint
expert

Code clean check

Detected a slowdown caused by scalar integer instructions (typically used for address computation). By removing them, you can lower the cost of an iteration from 6.83 to 4.00 cycles (1.71x speedup).

Workaround

Try to reorganize arrays of structures to structures of arrays
Consider to permute loops (see vectorization gain report)

Vectorization

Your loop is not vectorized. Only 10% of vector register length is used (average across all SSE/AVX instructions). By vectorizing your loop, you can lower the cost of an iteration from 6.83 to 2.00 cycles (3.42x speedup).

Details

All SSE/AVX instructions are used in scalar version (process only one data element in vector registers). Since your execution units are vector units, only a vectorized loop can use their full power.

Workaround

Try another compiler or update/tune your current one:
- recompile with ffast-math (included in Ofast) to extend loop vectorization to FP reductions.
Remove inter-iterations dependences from your loop and make it unit-stride:
- If your arrays have 2 or more dimensions, check whether elements are accessed contiguously and, otherwise, try to permute loops accordingly: C storage order is row-major: for(i) for(j) a[j][i] = b[j][i]; (slow, non stride 1) => for(i) for(j) a[i][j] = b[i][j]; (fast, stride 1)
- If your loop streams arrays of structures (AoS), try to use structures of arrays instead (SoA): for(i) a[i].x = b[i].x; (slow, non stride 1) => for(i) a.x[i] = b.x[i]; (fast, stride 1)

Execution units bottlenecks

Found no such bottlenecks but see expert reports for more complex bottlenecks.

FMA

Detected 3 FMA (fused multiply-add) operations.

Type of elements and instruction set

4 SSE or AVX instructions are processing arithmetic or math operations on double precision FP elements in scalar mode (one at a time).

Matching between your loop (in the source code) and the binary loop

The binary loop is composed of 7 FP arithmetical operations:

3: addition or subtraction (all inside FMA instructions)
3: multiply (all inside FMA instructions)
1: divide

The binary loop is loading 104 bytes (13 double precision FP elements). The binary loop is storing 24 bytes (3 double precision FP elements).

Arithmetic intensity

Arithmetic intensity is 0.05 FP operations per loaded or stored byte.

General properties

nb instructions	37
nb uops	37
loop length	178
used x86 registers	13
used mmx registers	0
used xmm registers	3
used ymm registers	0
used zmm registers	0
nb stack references	0

Front-end

MACRO FUSION NOT POSSIBLE FIT IN UOP CACHE

micro-operation queue	6.83 cycles
front end	6.83 cycles

Back-end

	P0	P1	P2	P3	P4	P5	P6	P7	P8	P9	P10	P11
uops	3.50	3.00	4.67	4.67	1.50	3.00	3.50	1.50	1.50	1.50	3.00	4.67
cycles	3.50	3.00	4.67	4.67	1.50	3.00	3.50	1.50	1.50	1.50	3.00	4.67

Execution ports to units layout:

P0 (256 bits): VPU, BRU, ALU, DIV/SQRT
P1 (256 bits): ALU, VPU
P2 (512 bits): load
P3 (512 bits): load
P4 (512 bits): store data
P5 (512 bits): ALU, VPU
P6: ALU, BRU
P7: store address
P8: store address
P9 (512 bits): store data
P10: ALU
P11 (512 bits): load

Cycles executing div or sqrt instructions

4.00

Cycles summary

Front-end	6.83
Dispatch	4.67
DIV/SQRT	4.00
Overall L1	6.83

Vectorization ratios

INT

all	0%
load	NA (no load vectorizable/vectorized instructions)
store	NA (no store vectorizable/vectorized instructions)
mul	NA (no mul vectorizable/vectorized instructions)
add-sub	0%
fma	NA (no fma vectorizable/vectorized instructions)
other	0%

all	0%
load	0%
store	0%
mul	NA (no mul vectorizable/vectorized instructions)
add-sub	NA (no add-sub vectorizable/vectorized instructions)
fma	0%
div/sqrt	0%
other	NA (no other vectorizable/vectorized instructions)

INT+FP

all	0%
load	0%
store	0%
mul	NA (no mul vectorizable/vectorized instructions)
add-sub	0%
fma	0%
div/sqrt	0%
other	0%

Vector efficiency ratios

INT

all	7%
load	NA (no load vectorizable/vectorized instructions)
store	NA (no store vectorizable/vectorized instructions)
mul	NA (no mul vectorizable/vectorized instructions)
add-sub	6%
fma	NA (no fma vectorizable/vectorized instructions)
other	7%

all	12%
load	12%
store	12%
mul	NA (no mul vectorizable/vectorized instructions)
add-sub	NA (no add-sub vectorizable/vectorized instructions)
fma	12%
div/sqrt	12%
other	NA (no other vectorizable/vectorized instructions)

INT+FP

all	10%
load	12%
store	12%
mul	NA (no mul vectorizable/vectorized instructions)
add-sub	6%
fma	12%
div/sqrt	12%
other	7%

Cycles and memory resources usage

Assuming all data fit into the L1 cache, each iteration of the binary loop takes 6.83 cycles. At this rate:

7% of peak load performance is reached (15.22 out of 192.00 bytes loaded per cycle (GB/s @ 1GHz))
2% of peak store performance is reached (3.51 out of 128.00 bytes stored per cycle (GB/s @ 1GHz))

Front-end bottlenecks

Performance is limited by instruction throughput (loading/decoding program instructions to execution core) (front-end is a bottleneck). By removing all these bottlenecks, you can lower the cost of an iteration from 6.83 to 4.67 cycles (1.46x speedup).

ASM code

In the binary file, the address of the loop is: 40ec5b

Instruction	Nb FU	P0	P1	P2	P3	P4	P5	P6	P7	P8	P9	P10	P11	Latency	Recip. throughput	Vectorization
INC %RAX	1	0	0	0	0	0	0	0	0	0	0	0	0	1	0.17	N/A
ADD $0x40,%ESI	1	0.20	0.20	0	0	0	0.20	0.20	0	0	0	0.20	0	1	0.20	scal (6.3%)
CMP %RCX,%RAX	1	0.20	0.20	0	0	0	0.20	0.20	0	0	0	0.20	0	1	0.20	scal (12.5%)
JE 40ec1b <advancePosition.extracted+0x6b>	1	0.50	0	0	0	0	0	0.50	0	0	0	0	0	0	0.50	N/A
MOV (%RDX,%RAX,4),%R11D	1	0	0	0.33	0.33	0	0	0	0	0	0	0	0.33	1	0.33	N/A
TEST %R11D,%R11D	1	0.20	0.20	0	0	0	0.20	0.20	0	0	0	0.20	0	2	0.20	scal (6.3%)
JLE 40ec50 <advancePosition.extracted+0xa0>	1	0.50	0	0	0	0	0	0.50	0	0	0	0	0	0	0.50	N/A
MOV %ESI,%R9D	1	0	0	0	0	0	0	0	0	0	0	0	0	1	0.17	N/A
MOV 0x20(%RBX),%R8	1	0	0	0.33	0.33	0	0	0	0	0	0	0	0.33	1	0.33	N/A
MOV 0x28(%RBX),%R10	1	0	0	0.33	0.33	0	0	0	0	0	0	0	0.33	1	0.33	N/A
MOV 0x10(%R8),%R15	1	0	0	0.33	0.33	0	0	0	0	0	0	0	0.33	1	0.33	N/A
MOV 0x18(%R8),%RDI	1	0	0	0.33	0.33	0	0	0	0	0	0	0	0.33	1	0.33	N/A
MOV 0x20(%R8),%R8	1	0	0	0.33	0.33	0	0	0	0	0	0	0	0.33	1	0.33	N/A
CMP $0x1,%R11D	1	0.20	0.20	0	0	0	0.20	0.20	0	0	0	0.20	0	1	0.20	scal (6.3%)
JE 40ed4c <advancePosition.extracted+0x19c>	1	0.50	0	0	0	0	0	0.50	0	0	0	0	0	0	0.50	N/A
LEA (%R9,%R9,2),%R12	1	0	0	0	0	0	0	0	0	0	0	0	0	0	0.17	N/A
LEA 0x28(,%R12,8),%R12	1	0	0	0	0	0	0	0	0	0	0	0	0	0	0.17	N/A
MOV %R11D,%R13D	1	0	0	0	0	0	0	0	0	0	0	0	0	1	0.17	scal (6.3%)
AND $0x7ffffffe,%R13D	1	0.20	0.20	0	0	0	0.20	0.20	0	0	0	0.20	0	1-2	0.20	scal (6.3%)
NOPL (%RAX,%RAX,1)	1	0	0	0	0	0	0	0	0	0	0	0	0	0	0.17	N/A
TEST $0x1,%R11B	1	0.20	0.20	0	0	0	0.20	0.20	0	0	0	0.20	0	2	0.20	N/A
JE 40ec50 <advancePosition.extracted+0xa0>	1	0.50	0	0	0	0	0	0.50	0	0	0	0	0	0	0.50	N/A
MOVSXD (%R15,%R9,4),%R11	1	0	0	0.33	0.33	0	0	0	0	0	0	0	0.33	1	0.33	N/A
SAL $0x4,%R11	1	0.50	0	0	0	0	0	0.50	0	0	0	0	0	0-2	0.50	N/A
SAL $0x3,%R9	1	0.50	0	0	0	0	0	0.50	0	0	0	0	0	0-2	0.50	N/A
LEA (%R9,%R9,2),%R9	1	0	0	0	0	0	0	0	0	0	0	0	0	0	0.17	N/A
VMOVSD (%R8,%R9,1),%XMM1	1	0	0	0.33	0.33	0	0	0	0	0	0	0	0.33	1	0.33	scal (12.5%)
VDIVSD 0x8(%R10,%R11,1),%XMM0,%XMM2	1	1	0	0.33	0.33	0	0	0	0	0	0	0	0.33	13-15	4	scal (12.5%)
VFMADD213SD (%RDI,%R9,1),%XMM2,%XMM1	1	0.50	0.50	0.33	0.33	0	0	0	0	0	0	0	0.33	4	0.50	scal (12.5%)
VMOVSD %XMM1,(%RDI,%R9,1)	1	0	0	0	0	0.50	0	0	0.50	0.50	0.50	0	0	1	0.50	scal (12.5%)
VMOVSD 0x8(%R8,%R9,1),%XMM1	1	0	0	0.33	0.33	0	0	0	0	0	0	0	0.33	1	0.33	scal (12.5%)
VFMADD213SD 0x8(%RDI,%R9,1),%XMM2,%XMM1	1	0.50	0.50	0.33	0.33	0	0	0	0	0	0	0	0.33	4	0.50	scal (12.5%)
VMOVSD %XMM1,0x8(%RDI,%R9,1)	1	0	0	0	0	0.50	0	0	0.50	0.50	0.50	0	0	1	0.50	scal (12.5%)
VMOVSD 0x10(%R8,%R9,1),%XMM1	1	0	0	0.33	0.33	0	0	0	0	0	0	0	0.33	1	0.33	scal (12.5%)
VFMADD213SD 0x10(%RDI,%R9,1),%XMM2,%XMM1	1	0.50	0.50	0.33	0.33	0	0	0	0	0	0	0	0.33	4	0.50	scal (12.5%)
VMOVSD %XMM1,0x10(%RDI,%R9,1)	1	0	0	0	0	0.50	0	0	0.50	0.50	0.50	0	0	1	0.50	scal (12.5%)
JMP 40ec50 <advancePosition.extracted+0xa0>	1	0	0	0	0	0	0	0	0	0	0	0	0	0	2.08	N/A

Function	advancePosition.extracted
Source file and lines	timestep.c:86-96
Module	exec

The loop is defined in /home/eoseret/qaas_runs_GNR/173-814-1858/intel/CoMD/build/CoMD/CoMD/src-openmp/timestep.c:86-96.

Warnings:

Non-innermost loop: analyzing only self part (ignoring child loops).
Ignoring paths for analysis
Too many paths. Rerun with max-paths=5
RecMII not computed since number of paths is unknown or > max_paths
Streams not analyzed since number of paths is unknown or > max_paths

hoisting them (moving them outside the loop)
turning them into conditional moves, MIN or MAX

Ex: if (x<0) x=0 => x = (x<0 ? 0 : x) (or MAX(0,x) after defining the corresponding macro)

gain
potential
hint
expert

Code clean check

Detected a slowdown caused by scalar integer instructions (typically used for address computation). By removing them, you can lower the cost of an iteration from 6.83 to 4.00 cycles (1.71x speedup).

Workaround

Try to reorganize arrays of structures to structures of arrays
Consider to permute loops (see vectorization gain report)

Vectorization

Details

All SSE/AVX instructions are used in scalar version (process only one data element in vector registers). Since your execution units are vector units, only a vectorized loop can use their full power.

Workaround

Try another compiler or update/tune your current one:
- recompile with ffast-math (included in Ofast) to extend loop vectorization to FP reductions.
Remove inter-iterations dependences from your loop and make it unit-stride:
- If your arrays have 2 or more dimensions, check whether elements are accessed contiguously and, otherwise, try to permute loops accordingly: C storage order is row-major: for(i) for(j) a[j][i] = b[j][i]; (slow, non stride 1) => for(i) for(j) a[i][j] = b[i][j]; (fast, stride 1)
- If your loop streams arrays of structures (AoS), try to use structures of arrays instead (SoA): for(i) a[i].x = b[i].x; (slow, non stride 1) => for(i) a.x[i] = b.x[i]; (fast, stride 1)

Execution units bottlenecks

Found no such bottlenecks but see expert reports for more complex bottlenecks.

FMA

Detected 3 FMA (fused multiply-add) operations.

Type of elements and instruction set

4 SSE or AVX instructions are processing arithmetic or math operations on double precision FP elements in scalar mode (one at a time).

Matching between your loop (in the source code) and the binary loop

The binary loop is composed of 7 FP arithmetical operations:

3: addition or subtraction (all inside FMA instructions)
3: multiply (all inside FMA instructions)
1: divide

The binary loop is loading 104 bytes (13 double precision FP elements). The binary loop is storing 24 bytes (3 double precision FP elements).

Arithmetic intensity

Arithmetic intensity is 0.05 FP operations per loaded or stored byte.

General properties

nb instructions	37
nb uops	37
loop length	178
used x86 registers	13
used mmx registers	0
used xmm registers	3
used ymm registers	0
used zmm registers	0
nb stack references	0

Front-end

MACRO FUSION NOT POSSIBLE FIT IN UOP CACHE

micro-operation queue	6.83 cycles
front end	6.83 cycles

Back-end

	P0	P1	P2	P3	P4	P5	P6	P7	P8	P9	P10	P11
uops	3.50	3.00	4.67	4.67	1.50	3.00	3.50	1.50	1.50	1.50	3.00	4.67
cycles	3.50	3.00	4.67	4.67	1.50	3.00	3.50	1.50	1.50	1.50	3.00	4.67

Execution ports to units layout:

P0 (256 bits): VPU, BRU, ALU, DIV/SQRT
P1 (256 bits): ALU, VPU
P2 (512 bits): load
P3 (512 bits): load
P4 (512 bits): store data
P5 (512 bits): ALU, VPU
P6: ALU, BRU
P7: store address
P8: store address
P9 (512 bits): store data
P10: ALU
P11 (512 bits): load

Cycles executing div or sqrt instructions

4.00

Cycles summary

Front-end	6.83
Dispatch	4.67
DIV/SQRT	4.00
Overall L1	6.83

Vectorization ratios

INT

all	0%
load	NA (no load vectorizable/vectorized instructions)
store	NA (no store vectorizable/vectorized instructions)
mul	NA (no mul vectorizable/vectorized instructions)
add-sub	0%
fma	NA (no fma vectorizable/vectorized instructions)
other	0%

all	0%
load	0%
store	0%
mul	NA (no mul vectorizable/vectorized instructions)
add-sub	NA (no add-sub vectorizable/vectorized instructions)
fma	0%
div/sqrt	0%
other	NA (no other vectorizable/vectorized instructions)

INT+FP

all	0%
load	0%
store	0%
mul	NA (no mul vectorizable/vectorized instructions)
add-sub	0%
fma	0%
div/sqrt	0%
other	0%

Vector efficiency ratios

INT

all	7%
load	NA (no load vectorizable/vectorized instructions)
store	NA (no store vectorizable/vectorized instructions)
mul	NA (no mul vectorizable/vectorized instructions)
add-sub	6%
fma	NA (no fma vectorizable/vectorized instructions)
other	7%

all	12%
load	12%
store	12%
mul	NA (no mul vectorizable/vectorized instructions)
add-sub	NA (no add-sub vectorizable/vectorized instructions)
fma	12%
div/sqrt	12%
other	NA (no other vectorizable/vectorized instructions)

INT+FP

all	10%
load	12%
store	12%
mul	NA (no mul vectorizable/vectorized instructions)
add-sub	6%
fma	12%
div/sqrt	12%
other	7%

Cycles and memory resources usage

Assuming all data fit into the L1 cache, each iteration of the binary loop takes 6.83 cycles. At this rate:

7% of peak load performance is reached (15.22 out of 192.00 bytes loaded per cycle (GB/s @ 1GHz))
2% of peak store performance is reached (3.51 out of 128.00 bytes stored per cycle (GB/s @ 1GHz))

Front-end bottlenecks

ASM code

In the binary file, the address of the loop is: 40ec5b

Instruction	Nb FU	P0	P1	P2	P3	P4	P5	P6	P7	P8	P9	P10	P11	Latency	Recip. throughput	Vectorization
INC %RAX	1	0	0	0	0	0	0	0	0	0	0	0	0	1	0.17	N/A
ADD $0x40,%ESI	1	0.20	0.20	0	0	0	0.20	0.20	0	0	0	0.20	0	1	0.20	scal (6.3%)
CMP %RCX,%RAX	1	0.20	0.20	0	0	0	0.20	0.20	0	0	0	0.20	0	1	0.20	scal (12.5%)
JE 40ec1b <advancePosition.extracted+0x6b>	1	0.50	0	0	0	0	0	0.50	0	0	0	0	0	0	0.50	N/A
MOV (%RDX,%RAX,4),%R11D	1	0	0	0.33	0.33	0	0	0	0	0	0	0	0.33	1	0.33	N/A
TEST %R11D,%R11D	1	0.20	0.20	0	0	0	0.20	0.20	0	0	0	0.20	0	2	0.20	scal (6.3%)
JLE 40ec50 <advancePosition.extracted+0xa0>	1	0.50	0	0	0	0	0	0.50	0	0	0	0	0	0	0.50	N/A
MOV %ESI,%R9D	1	0	0	0	0	0	0	0	0	0	0	0	0	1	0.17	N/A
MOV 0x20(%RBX),%R8	1	0	0	0.33	0.33	0	0	0	0	0	0	0	0.33	1	0.33	N/A
MOV 0x28(%RBX),%R10	1	0	0	0.33	0.33	0	0	0	0	0	0	0	0.33	1	0.33	N/A
MOV 0x10(%R8),%R15	1	0	0	0.33	0.33	0	0	0	0	0	0	0	0.33	1	0.33	N/A
MOV 0x18(%R8),%RDI	1	0	0	0.33	0.33	0	0	0	0	0	0	0	0.33	1	0.33	N/A
MOV 0x20(%R8),%R8	1	0	0	0.33	0.33	0	0	0	0	0	0	0	0.33	1	0.33	N/A
CMP $0x1,%R11D	1	0.20	0.20	0	0	0	0.20	0.20	0	0	0	0.20	0	1	0.20	scal (6.3%)
JE 40ed4c <advancePosition.extracted+0x19c>	1	0.50	0	0	0	0	0	0.50	0	0	0	0	0	0	0.50	N/A
LEA (%R9,%R9,2),%R12	1	0	0	0	0	0	0	0	0	0	0	0	0	0	0.17	N/A
LEA 0x28(,%R12,8),%R12	1	0	0	0	0	0	0	0	0	0	0	0	0	0	0.17	N/A
MOV %R11D,%R13D	1	0	0	0	0	0	0	0	0	0	0	0	0	1	0.17	scal (6.3%)
AND $0x7ffffffe,%R13D	1	0.20	0.20	0	0	0	0.20	0.20	0	0	0	0.20	0	1-2	0.20	scal (6.3%)
NOPL (%RAX,%RAX,1)	1	0	0	0	0	0	0	0	0	0	0	0	0	0	0.17	N/A
TEST $0x1,%R11B	1	0.20	0.20	0	0	0	0.20	0.20	0	0	0	0.20	0	2	0.20	N/A
JE 40ec50 <advancePosition.extracted+0xa0>	1	0.50	0	0	0	0	0	0.50	0	0	0	0	0	0	0.50	N/A
MOVSXD (%R15,%R9,4),%R11	1	0	0	0.33	0.33	0	0	0	0	0	0	0	0.33	1	0.33	N/A
SAL $0x4,%R11	1	0.50	0	0	0	0	0	0.50	0	0	0	0	0	0-2	0.50	N/A
SAL $0x3,%R9	1	0.50	0	0	0	0	0	0.50	0	0	0	0	0	0-2	0.50	N/A
LEA (%R9,%R9,2),%R9	1	0	0	0	0	0	0	0	0	0	0	0	0	0	0.17	N/A
VMOVSD (%R8,%R9,1),%XMM1	1	0	0	0.33	0.33	0	0	0	0	0	0	0	0.33	1	0.33	scal (12.5%)
VDIVSD 0x8(%R10,%R11,1),%XMM0,%XMM2	1	1	0	0.33	0.33	0	0	0	0	0	0	0	0.33	13-15	4	scal (12.5%)
VFMADD213SD (%RDI,%R9,1),%XMM2,%XMM1	1	0.50	0.50	0.33	0.33	0	0	0	0	0	0	0	0.33	4	0.50	scal (12.5%)
VMOVSD %XMM1,(%RDI,%R9,1)	1	0	0	0	0	0.50	0	0	0.50	0.50	0.50	0	0	1	0.50	scal (12.5%)
VMOVSD 0x8(%R8,%R9,1),%XMM1	1	0	0	0.33	0.33	0	0	0	0	0	0	0	0.33	1	0.33	scal (12.5%)
VFMADD213SD 0x8(%RDI,%R9,1),%XMM2,%XMM1	1	0.50	0.50	0.33	0.33	0	0	0	0	0	0	0	0.33	4	0.50	scal (12.5%)
VMOVSD %XMM1,0x8(%RDI,%R9,1)	1	0	0	0	0	0.50	0	0	0.50	0.50	0.50	0	0	1	0.50	scal (12.5%)
VMOVSD 0x10(%R8,%R9,1),%XMM1	1	0	0	0.33	0.33	0	0	0	0	0	0	0	0.33	1	0.33	scal (12.5%)
VFMADD213SD 0x10(%RDI,%R9,1),%XMM2,%XMM1	1	0.50	0.50	0.33	0.33	0	0	0	0	0	0	0	0.33	4	0.50	scal (12.5%)
VMOVSD %XMM1,0x10(%RDI,%R9,1)	1	0	0	0	0	0.50	0	0	0.50	0.50	0.50	0	0	1	0.50	scal (12.5%)
JMP 40ec50 <advancePosition.extracted+0xa0>	1	0	0	0	0	0	0	0	0	0	0	0	0	0	2.08	N/A

Report Configuration

Code clean check

Workaround

Vectorization

Details

Workaround

Execution units bottlenecks

FMA

Type of elements and instruction set

Matching between your loop (in the source code) and the binary loop

Arithmetic intensity

General properties

Front-end

Back-end

Cycles summary

Vectorization ratios

Vector efficiency ratios

Cycles and memory resources usage

Front-end bottlenecks

ASM code

Code clean check

Workaround

Vectorization

Details

Workaround

Execution units bottlenecks

FMA

Type of elements and instruction set

Matching between your loop (in the source code) and the binary loop

Arithmetic intensity

General properties

Front-end

Back-end

Cycles summary

Vectorization ratios

Vector efficiency ratios

Cycles and memory resources usage

Front-end bottlenecks

ASM code